Loading video...

Video Failed to Load

Go Home

VS Code was already used by millions of developers for agentic coding. However, the editor layout has traditionally been optimized for single-task and single-workspace workflows. Today, we're introducing a new window to enable our users (and ourselves!) to work with multiple agents across multiple projects: Agents. Now available in...

151,517 views • 4 months ago •via X (Twitter)

42 Comments

Pierce Boggan's profile picture
Pierce Boggan4 months ago

Agents window reflects how our own workflows on the VS Code team evolved to focus more on higher-level partnership with the agent. A year ago, many of us used to spend an immense amount of time following the specific agent trajectory, including the tool calls and edits it made. Now, we spend our time building alignment with the models with features like Plan mode, and reviewing outcomes as a first step rather than code.

Pierce Boggan's profile picture
Pierce Boggan4 months ago

The Agents window also expands our vision with Agent HQ to not just give you choice in state of the art models, but also harnesses as well. Give our Claude agent integration a try, all powered by your GitHub Copilot subscription:

Pierce Boggan's profile picture
Pierce Boggan4 months ago

We also want to be able to continue our work from anywhere. The Agents window also includes support for remote development, allowing you to connect to coding agents running in remote hosts, such as a cloud sandbox.

Pierce Boggan's profile picture
Pierce Boggan4 months ago

We also built a brand new experience to allow you to work from the browser (including mobile!) with your coding agents!

Stephen Turner 🇬🇧🇺🇦's profile picture
Stephen Turner 🇬🇧🇺🇦4 months ago

I look forward to the Codex harness being included. This is the way.

Abram Jackson's profile picture
Abram Jackson4 months ago

Terrific! Do you use the words "Agents" and "sessions" interchangeably? Or is an "agent" a combination of repo with its skills and enabled MCP Servers?

Pierce Boggan's profile picture
Pierce Boggan4 months ago

Nope, though sometimes I accidentally do :D Agent = core agent loop = the prompt, tools, context. In VS Code, we support multiple agent types, including VS Code (Local), Copilot CLI, Claude Session = An instance of that agent in a particular repo/workspace

lioncubs's profile picture
lioncubs3 months ago

@burkeholland @pierceboggan this app shows NO Love to the ultimate Remote Session WSL - what about sessions in devcontainers as well??? Also another awesome “Remote Session” 😭

Pierce Boggan's profile picture
Pierce Boggan3 months ago

@burkeholland Adding @BrigitMurtaugh for Agents app + remote development

Jun(e)'s profile picture
Jun(e)4 months ago

any plan to support TUI ? actually I mean the support for opencode / codex / claude code

Prismor's profile picture
Prismor4 months ago

We love VS code! Thanks for shipping this

AI Inc's profile picture
AI Inc4 months ago

Nice feature, but I cancelled my GitHub Copilot Pro+ account now, because of the high prices coming up... Good luck!

CapyToolkit's profile picture
CapyToolkit4 months ago

Thanks for shipping this @pierceboggan! VS Code rocks 🤘🏻

Nilay's profile picture
Nilay4 months ago

Hey quick question, is there information to access, why suddenly my locally hosted models and BYOMs (from Foundry etc.) are no longer visible and accessible via VS Code Chat or Agents? Is there a shift?

Ami Schreiber's profile picture
Ami Schreiber4 months ago

It’s the ADHD window!

tianma's profile picture
tianma4 months ago

只能使用copilot的模型,我不会去用,我使用codex,claude价格公道。

lifcc's profile picture
lifcc4 months ago

Agents window is the right primitive. The pain is not launching another agent; it is keeping repo state, diffs, terminal output, and handoff visible when two projects are moving at once.

armandoki's profile picture
armandoki4 months ago

I’ve been using it a lot the past few weeks. 😎

Pierce Boggan's profile picture
Pierce Boggan4 months ago

What do you think?

armandoki's profile picture
armandoki3 months ago

I'm so glad we're building this! Lots of thoughts. Will ping you.

Abdi Daud's profile picture
Abdi Daud3 months ago

Well done team; I've been using it for sometime glad to see it in stable. Do you have a plan for scheduled tasks say that run on cloud and/or PC

David Dossett's profile picture
David Dossett3 months ago

congrats on the 🚢!

Eshetie ፼ Liku's profile picture
Eshetie ፼ Liku4 months ago

#coolstorybro I ran out of premium requests from my copilot pro+ subscription in the first week, all(most?) requests are user initiated. The allotment does not match the IDE/AI work @code is making. Difficult position to be in for me.

Sant's profile picture
Sant4 months ago

love visual studio code for agentic development much better that anything else out there. but multiple agents across multiple projects... sounds like software apocalypse.

Daniel's profile picture
Daniel4 months ago

Too bad no one can afford to do that now.

spencer's profile picture
spencer3 months ago

wicked

Niclas's profile picture
Niclas4 months ago

Remote SSH sessions will not work/break for me ...

Aravind Putrevu's profile picture
Aravind Putrevu3 months ago

This looks like superset!

RAFA™'s profile picture
RAFA™3 months ago

@sseraphini 👏🏽👏🏽

Egor Egorov's profile picture
Egor Egorov3 months ago

@film_girl Is actual programming still supported? I use vscode since Atom inception. It would be kind of sad to update one day and see code editor gone.

Pierce Boggan's profile picture
Pierce Boggan3 months ago

@film_girl 💯, editor isn't going anywhere! We are landing improvements to that experience too. New markdown experience just shipped yesterday!

Egor Egorov's profile picture
Egor Egorov3 months ago

@film_girl Oh yes, I was pleasantly surprised with that yesterday! These days it was unexpected. It is actually what I was missing from vs code and thus it is a very welcome addition!

Filip Mares's profile picture
Filip Mares3 months ago

It's my daily. Love it!

Prompt Heat's profile picture
Prompt Heat3 months ago

WTF? i cant buy other plans , this is bullshit

Alexander Rehm's profile picture
Alexander Rehm3 months ago

Odd bug, with @claudeai windows I'm only ever seeing Haiku as the model, despite the existing sessions being Sonnet or Opus. Known issue?

Iulian Lita's profile picture
Iulian Lita3 months ago

Doesn't really matter. The upcoming copilot changes make all this kinda useless. Let's wait a few months to see what will change after 20-30%+ of the users will cancel their copilot sub.

Tanay Salpekar's profile picture
Tanay Salpekar3 months ago

The app looks / feels great. The only thing: I’m not trying to pay API usage rates with 3rd party models.

Tech With Matteo's profile picture
Tech With Matteo4 months ago

running multiple agents across projects at once is exactly where the real productivity gains are, the single-task editor felt like a bottleneck once you get used to parallel work

Rennan Felizardo's profile picture
Rennan Felizardo4 months ago

it's indeed a good app, but unfortunately I stopped using it at the company months ago due to the lack of GHE cloud support. Session delegation to copilot cli doesn't work when the host is not the default one. Which is quite odd since GitHub and VS Code are backed by Microsoft

Sergio Pereira's profile picture
Sergio Pereira3 months ago

Have you been able to manage what happens when two of those agents edit the same file? Worktrees isn't the solution....

Anthony Humphreys's profile picture
Anthony Humphreys4 months ago

@JamesMontemagno Need more sidebars @pierceboggan !

Gaurang Karia's profile picture
Gaurang Karia3 months ago

Looks great.. looking forward to using it.

Related Videos

New short course: Building Code Agents with Hugging Face smolagents! Learn how to build code agents in this course, created in collaboration with Hugging Face, and taught by Thomas Wolf, its co-founder and CSO, and m_ric, Hugging Face’s Project Lead on Agents. Tool-calling agents use LLMs to generate multiple function calls sequentially to complete a complex sequence of tasks. They generate one function call, execute it, observe, reason, and decide what to do next. Code agents take a different approach. They consolidate all these calls into a single block of code, letting the LLM lay out an entire action plan at once, which can be executed efficiently to provide more reliable results. You’ll learn how to code agents using smolagents, a lightweight agentic framework from Hugging Face. Along the way, you’ll learn how to run LLM-generated code safely and develop an evaluation system to optimize your code agent for production. In detail, you’ll learn: - How agentic systems have evolved, gaining greater levels of agency over time—and why code agents are a next step. - How code agents write their actions in code. - When code agents outperform function-calling agents. - How to run code agents safely in your system using a constrained Python interpreter and sandboxing using E2B. - To trace, debug, and assess the code agent to optimize its behaviours for complex requests. - How to build a research multi-agent system that can find information online and organize it into an interactive report. By the end of this course, you’ll know how to build and run code agents using smolagents, and deploy them safely with a structured evaluation system in your projects. Please sign up here!

Andrew Ng

127,724 views • 1 year ago